Let's jump right in. At 2AM on a Tuesday, my team launched a huge feature release for an enterprise publishing platform. The launch was perfect.I went to sleep thinking it was a success. But at 8AM the next morning, my phone started to go off. As our editors start their morning shift, the whole platform went haywire. My engineering team was in full fire firefighting mode trying to figure out what went wrong, but there was no bug. The code was doing exactly what we built it to do. The problem was invisible to us. Our editors were under tight morning deadlines. When they uploaded high res images, the new workflow took a few seconds to process. We didn't give them a loading screen or any feedback. We accidentally overlooked this.

So they thought it was broken. They went what they used to they did what they used to do. They just keep clicking upload over and over again. These human error processes queued up thousands of redundant image processing jobs and database tests at once. The database locked up and the platform crashed. The technology didn't fail, our alignment failed. Our developers didn't understand the editor's rush, and our our editors didn't understand the new workflow speed. Because one group didn't understand how the new system worked, the entire platform paid the price. I think about that morning a lot because today, I see that exact same pattern happening everywhere, and it is happening at a massive scale. Almost every company uses AI right now, but very few get the result they expect.

Mackenzie data shows that 88% of companies use AI today, but only about 6% are actually high performers. The winners didn't buy better tools. They redesigned how their teams work together with it. Because when we add speed without team alignment, we don't get innovation. We get fragile systems. Let's look at the let's look at the reality of AI in engineering right now. Developers are using AI to write code 34% faster. This sounds great, right? But in reality, bugs are up 54%. The incident to pull request ratio has skyrocketed by 242%, and code churn is up an incredible 861%. Automation takes away manual work, but it dramatically increases the chance of a massive system crash and failure. Let me give you a two real world example of what this fragility looks like. Some of you may remember this. In November 2025, Cloudflare had a massive outage.

It took down 20% of global Internet traffic. It took down ChatGPT, Spotify, and Shopify. What happened? What caused this was that an automated feature hit a hard coded limit of 200 items. The automated the automation moves so fast, the system collapsed under its own weight. Or look at July 2025, when an AI agent at a major tech company was supposed to be doing routine work during a code freeze. Instead, it accidentally deleted a live production database. It wiped out over 1,000 executive records and then fabricated 4,000 fake users to replace them. The speed of AI is incredible, but without human oversight, the blast radius is terrifying. So what are the top 6% of the companies doing differently to prevent this? A landmark Harvard study looked at over 700 professionals at Procter and Gamble. They found that AI enhanced cross functional teams were three times more likely to create breakthrough solutions.

AI acted as a bridge between their different specialties. Teloy's 2026 data backs this up. High performing AI teams report dramatically better collaboration, 79% compared to just 57% for low performing teams. The secret isn't technology. It's a trust and alignment. To build that alignment, we use a model called the three e framework. It looks at three experiences where you perform your your platform either holds together or falls apart. The audience experience, the people using the product, the people you build for, The creator experience, the people who create and manage the content and operations. The developer experience, the people building the code and underlying architecture. When these three experiences are aligned and leveraging AI on top, AI makes you unstoppable. When they aren't, the crack the cracks start to show. And right now, those cracks are getting wider.

A recent Atlassian study found that sixty three percent of developers say that their leaders don't understand their daily pain points. That is a massive empathy gap, and it is exactly where platform failures begin. So let's apply this three e framework to the future. Let's look at your future engineering bench. The people who run our platform in five to ten years, what happens if they're they're not properly trained or equipped on these new tools and technology. First, audience experience. If the people building the app don't understand AI, you get huge blind spots in what you build for the users. Second, the creator experience. If your editors aren't taught AI literacy, workflows break down at the handoff. That is exactly how you get the massive database crash I told you about at the beginning.

Finally, the developer experience. As you saw, AI lets senior engineers write write code faster than ever. But then speed is flooding our system. Ferro's data showed that code merged without a human review has spiked by 31.3%. The safety checks are failing. Someone has to review the code. Someone has to catch what AI gets wrong. But if half of your junior teams isn't one being actively mentored or trained, or taught how to use AI or can't read the code that AI generated, they can't catch these errors. You're building technical depth directly into your pipeline. So we have to ask, who's actually being trained to catch these errors? I look at the data to see who's who's getting the support to be properly trained.

I expect it to find normal training gaps, like, maybe the difference between juniors and senior staff or difference different engineering teams. But when I looked closer, I found something interesting, rather surprising. Something I didn't think of initially. The data rebuild, a massive structure flaw, and who's actually adopting these tools and technology. When I looked at the numbers, men are currently 22% more likely to use AI daily at work, and that gap isn't closing. It's getting wider. Oxford researchers found that the personal use AI gender gap grew from 3.7 to 5.3 points just over the last year. Now this isn't just a minor shift in habits. It is massive risk to our talent pipeline. You might also wonder if this is just an access problem or a training problem. It isn't.

Researchers looked at 17,000 entrepreneurs in Kenya. They gave everyone the exact same AI tools. They gave everyone the exact same training. Even with AI even with a, perfectly level playing field, women were still 13% less likely to try the AI. Why? A massive Harvard study of 140,000 people gives us the answer. Women worry more about the ethics of these tools and technology. They fear looking like they're cheating. And most importantly, women face much harsher professional penalties if they make a mistake. I saw this happening with women entrepreneurs that women entrepreneurs as well. When they pitch to VCs, they don't just walk in with a half baked idea. They start slower. They research meticulously. They make sure they have the back data to speak with absolute bulletproof confidence because they know they're judged more carefully.

Women in tech also know that they're judged more carefully. So they research more and longer. They wait until they're absolutely sure. That exact same careful, fully researched habit is playing out right now with AI adoption. Because women are smartly waiting for clear permissions, we have to look at who's actually getting the clear permissions. At the entry level, the future bench we just talked about, only 21% of women get explicit manager encouragement to use AI. Compared to for men, it's thirty three percent. Now if you have been in technology industry for a while, you might be thinking, why do they need encouragement? It's just technology. Open the tools and figure it out. But remember those harsher penalties? In this case, manager encouragement is not cheerleading. It is operational cover.

It is managers explicitly saying, I want you to test these tools and technology on company time, and if it messes up or hallucinates, I have your back. Harvard found that giving people this psychological safety is a single biggest lever we have to close the gap. If you don't fix this, we aren't just failing to be inclusive. We're leaving a massive weak spot in our talent pipeline. The good news is we can fix this, and it doesn't take a massive corporate budget or six months HR plan. Here are three highly practical steps you can take starting next week. One, run a capacity audit. Don't just look at your software licenses to see who has access to AI. Look at your look at your team's calendars.

Who's actually getting the block time and mentoring to master it? If you find that this training correlates with gender, seniority, or with specific team assignments, you have just found your risk map. This is exactly where your next platform failure is hiding. Give everyone the structured time they need, then you protect your talent pipeline. Action two, cross functional pairing, not workshops. A Friday lunch and learn checks the compliance boxes, but it doesn't build real skills. A pilot who only trains on autopilot is not the person you want landing your plane in a storm. Instead, set up a structured pairing. Put your AI experts next to people who need training, and make sure you pair teams across different teams. Put developers with editors. Remember the Harvard study?

AI enhanced cross functional teams create three times more breakthroughs. Have them sit together and learn by doing. Action three, formalize operational cover. Right now, the data on AI rules is alarming. 78% of executives admit that they could not pass an AI governance audit today. Even worse, 49 of employees admit they're using hidden unsanctioned AI tools at work right now. Why? Because there are no rules, no clear rules. We have to stop seeing AI governance as a roadblock. Good rules give your team a safe harbor. You don't need a 100 page legal document. You just need a simple one page framework that gives your team the explicit, clear permission they need to experiment without fear. Give them the rules of the road, and the people most attuned to risk will finally feel safe enough to try and fly.

Think back to that eight AM database crash I told you about at the beginning. The code was perfect, the technology worked, but the platform still broke because the team wasn't aligned. The data shows that exact same pattern is happening right now everywhere at scale. The leaders who will carry our platforms into the next decade will be the ones who actively develop everyone on their bench, not the ones who raise their hands first. Provide that operational cover is not a value statement. It's an engineering decision that we can make. And right now, it is the most important one you can make. Thank you very much.